Moonshot's Kimi K3.1 Pro Scores 55 on AI Index

Moonshot AI has released Kimi K3.1 Pro, a new reasoning model that immediately claimed the fourth position on the Artificial Analysis Intelligence Index. The system achieved a score of 55 points, placing it just three points behind the leading proprietary models. This performance marks a significant leap for the Chinese developer in the global artificial intelligence race.

The new model represents a post-trained evolution of the previous K3 version, utilizing significantly more compute during the training phase. This upgrade resulted in an 11-point increase over its predecessor. Currently, only three models rank higher: Claude Opus 5.5 with 58 points, Claude Sonnet 5.5 with 56 points, and Muse Wonder Full also with 56 points. If Moonshot adheres to its plan to release the model weights by November 4, 2026, K3.1 Pro will become the highest-scoring open-weights model available. That move would narrow the performance gap between open and proprietary systems from 12 points to just three, assuming no other major releases occur in the interim.

Moonshot is also launching a variant named K3.1 Pro Highspeed, which the company states runs two to three times faster than the standard version. Both options are currently accessible through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. The most notable improvements in K3.1 Pro appear in agentic tasks. On the GDPval-AA benchmark, the model reached 1,846 Elo, a 312-point increase over K3. It now sits slightly behind Claude Opus 5.5 and Muse Wonder Full. In coding evaluations, Moonshot reports a 65.7 percent score on Terminal-Bench 4.0, surpassing GPT-6 Astra and closely trailing Claude Opus 5.5.

The model also demonstrated sharper accuracy in specific technical areas. Hallucination rates dropped from 23.4 percent to 8.3 percent on Moonshot's internal stress test, though GPT-6 Astra maintains a lower rate of 5.5 percent. In finance, the model improved by 7.2 points in modeling and 11.6 points in research. For 3D modeling, K3.1 Pro works directly with engines like Unity, Unreal Engine, and Godot. It scored 92.7 percent with tools on BenchCAD, a result close to the top proprietary systems. The company has also introduced a 3D Make feature on its website to allow users to preview and export models.

Despite the performance gains, the model retains the same architectural dimensions as K3, featuring 2.8 trillion total parameters with 104 billion active. It supports a one-million token context window and processes both text and images. According to Artificial Analysis, it sits at the top of the Pareto frontier for its size. Token prices have decreased by 33 percent, with input costs now at $2 per million tokens and output at $10. The cost per Intelligence Index task remains nearly unchanged at $1.98, as the model generates 14 percent more output tokens.

Moonshot also addressed cybersecurity capabilities in its technical release. The company noted that vulnerability research skills grew as post-training scaled up. To prepare for the open-weights release, Moonshot removed cyber-related data from the training phases. On an internal test involving recent software vulnerabilities, the uncleaned version solved 90.5 percent of tasks, while the final K3.1 Pro version solved 81.0 percent. This is higher than DeepSeek-V4.1-Flash at 85.7 percent, but Moonshot argues that open models should gradually raise the bar for defensive purposes rather than pushing the absolute limit of offensive capabilities.

The next major step for this model is the scheduled release of its weights next month. This event will determine if it successfully claims the top spot among open-weights competitors. The pricing changes and performance benchmarks suggest a competitive shift in the market, particularly for developers relying on cost-effective, high-performance reasoning models for coding and agentic workflows.